deberta-v3-large-survey-cross_passage_consistency-rater-gpt4
This model is a fine-tuned version of microsoft/deberta-v3-large on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4075
- Krippendorff: 0.7501
- Spearman: 0.8372
- Absolute Agreement: 0.8907
- Agreement Within One: 0.9799
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 6e-06
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Krippendorff | Spearman | Absolute Agreement | Agreement Within One |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 50 | 2.0452 | 0.0001 | nan | 0.0694 | 0.7778 |
| No log | 2.0 | 100 | 2.0337 | 0.0175 | 0.0377 | 0.0694 | 0.7778 |
| No log | 3.0 | 150 | 2.0273 | -0.0372 | -0.0377 | 0.0972 | 0.7778 |
| No log | 4.0 | 200 | 2.0939 | -0.1244 | -0.1615 | 0.125 | 0.7083 |
| No log | 5.0 | 250 | 2.1835 | -0.1887 | 0.2306 | 0.3194 | 0.6667 |
| No log | 6.0 | 300 | 2.1351 | -0.2704 | nan | 0.3194 | 0.6667 |
| No log | 7.0 | 350 | 2.2556 | -0.2704 | nan | 0.3194 | 0.6667 |
| No log | 8.0 | 400 | 2.2676 | -0.2704 | nan | 0.3194 | 0.6667 |
| No log | 9.0 | 450 | 2.3103 | -0.2704 | nan | 0.3194 | 0.6667 |
| 1.3588 | 10.0 | 500 | 2.2637 | -0.2704 | nan | 0.3194 | 0.6667 |
| 1.3588 | 11.0 | 550 | 2.3671 | -0.2354 | -0.1828 | 0.25 | 0.6806 |
| 1.3588 | 12.0 | 600 | 2.1781 | -0.2919 | -0.2073 | 0.2917 | 0.6667 |
| 1.3588 | 13.0 | 650 | 2.1614 | -0.1676 | 0.0113 | 0.2917 | 0.6806 |
| 1.3588 | 14.0 | 700 | 2.4461 | -0.1539 | -0.0635 | 0.2917 | 0.6806 |
| 1.3588 | 15.0 | 750 | 2.4148 | -0.0466 | 0.0219 | 0.3056 | 0.6944 |
| 1.3588 | 16.0 | 800 | 2.4528 | -0.0885 | -0.0518 | 0.2361 | 0.6806 |
| 1.3588 | 17.0 | 850 | 2.4096 | -0.0964 | 0.1186 | 0.3194 | 0.6806 |
| 1.3588 | 18.0 | 900 | 2.3058 | 0.1768 | 0.2017 | 0.3056 | 0.7639 |
| 1.3588 | 19.0 | 950 | 2.5967 | -0.1308 | -0.0542 | 0.3056 | 0.6944 |
| 0.6327 | 20.0 | 1000 | 3.1201 | -0.1148 | 0.0260 | 0.3194 | 0.6806 |
Framework versions
- Transformers 4.26.0
- Pytorch 1.13.1
- Datasets 2.10.1
- Tokenizers 0.12.1